Power Plan Controversy

UPDATE 8/2010: I wanted to bump this post for those who haven't seen this before. These tests still apply for Windows 7 as they did for Vista. Basically, setting your power plan to Balanced is better than High Performance 99% of the time


Since my laptop is 99% of the time plugged in, I have had my power plan set to High Performance since the beginning. Recently I was doing a little research regarding the 3 standard power plans, and have found some interesting results.

First, here is a basic description of the 3 power plans:

Like most people know, to get the best battery life switch to Power Saver (or create a custom plan).
And if you want to be in the middle between performance and battery life, switch to Balanced.
And to get the best performance, most assume that High Performance is best.
That's what I thought too, but according to some tests, this isn't necessarily true.

According to Jenn K. Lee from pocketables.net, her benchmarks showed that Balanced may be better than High Performance. But how could this be?

Here's her CrystalMark 2004R3 benchmark tests on an HTC Shift X9500:

Balanced Power Plan:


Power Saver Power Plan:


High Performance Power Plan:



According to this test, Balanced beats out High Performance.
----
Then she ran the same test on a Sony Vaio TZ170N.
Here's the results:




Once again, Balanced beats out High Performance.

For those interested, here are the links to her tests:
1. http://www.pocketables.net/2008/05/htc-shift-power.html
2. http://www.pocketables.net/2008/05/sony-vaio-tz-po.html



So if these tests are correct, we should be using Balanced over High Performance because we would get better performance and save power as well. I'm still skeptical though.

Some people think there is a flaw in CrystalMark2004R3, so maybe so other forum members here can run their own benchmarks and see if these test results are accurate or not.

So I hope this information helps, and feel free to post your own findings and tests. I will edit this post if I find any new info.

EDIT: Some more recent tests have shown a less than 2% difference between "Balanced" and "High Performance." So even if Balanced is not necessarily better performance than High Performance, it still is probably the better power plan to choose because you save power without any performance loss.
Note: The dots in the Windows Vista Power Options are misleading because it shows Balanced as half the performance of "High Performance". This is inaccurate. In Windows 7, these dots are no longer shown.

Tests are still be conducted so more info on the way


Consensus so far: Balanced is the best choice. It has basically the same performance of "High Performance" while also conserving power thus leading to a cooler computer.

 

balanced is just as fast as high performance, and due to the throttling down while idling, it may let your system cool down more. this might gain your performance on core i7 systems that can autoclock.

that's just one idea.

but one thing is quite logical: high performance is for the ones that want "high performance". but just because they like to read the word. balanced gets the SAME performance when needed. it just trothles down when not needed. and this is a GOOD thing. better cooling, more silent fans, more batterylife.

balanced > high performance.



edit: and btw, those profiles are not "magic". they are 100% editable, and documented. so you can reconfigure them to what ever needs you have (i do so), and maybe find out whats the difference there, in their settings.

 

I agree that Balanced makes more sense. Many people have set their Power Plan to High Performance without giving it much thought, so hopefully this thread will help people see that High performance may not be best.

 

i guess that's why high performance got hidden away in win7 because only balanced and energy saving are the ones that normal people ever should need.

 

This is informative and useful. Thanks a lot.

 

It is informative.
I always want the laptop to perform as fast as it capable all the time.
So, I used RMClock. To make it run at highest FID and lowest VID(without BSOD) all the time. It works better than "High performance" in the Control Panel Power Option. Save electricity(battery last longer due to lowest VID) with highest speed(max FID).
I never trust build-in Power Option in the Control Panel.

 

Glad to help


Just wondering...I know that the Balanced plan throttles the CPU when it's not needed at max performance, but does it also affect the GPU at all?

And I'm also still wondering why Balanced would result in better performance than High performance. Even if they were equal, that would be a good enough reason for people to rethink their current setting and switch to Balanced. But better yet I suppose, Balanced is actually even better at achieving Max performance..? Very interesting. I'll probably run some of my own tests to compare to the ones I already listed. This is a very interesting debate.
Plus I wonder how power plans like HP recommended compare to balanced...

 

A bit old, but give it a look.

Edit: Oopps, sorry, forgot the link Anyway, here it is: http://tekvax.com/dan/?q=node/14

 

More information I found that may be useful to this whole investigation.

These are the details of each power plan:

Power Plan — Balanced (Set as active plan by default)

* Require password on: Yes
* Turn off hard disks after: 20 minutes
* Wireless adapter settings — Power Saving Mode: Maximum Performance
* Sleep — Sleep after: 60 minutes
* Sleep — Hibernate after: Never
* USB Settings — USB Selective suspend setting: Disabled
* Power button action: Shut Down
* Start menu power button action: Sleep
* PCI Express — Link State Power Management: Moderate power savings
* Processor Power Management — Minimum Processor State — 5%
* Processor Power Management — Maximum Processor State — 100%
* Search and Indexing — Power savings mode: High Performance.
* Turn off the display: 20 minutes
* Adaptive display: On
* Multimedia settings — When sharing media: Prevent idling to sleep

Power Plan — Power Saver

* Require password on: Yes
* Turn off hard disks after: 20 minutes
* Wireless adapter settings — Power Saving Mode: Maximum Performance
* Sleep — Sleep after: 60 minutes
* Sleep — Hibernate after: Never
* USB Settings — USB Selective suspend setting: Disabled
* Power button action: Shut Down
* Start menu power button action: Sleep
* PCI Express — Link State Power Management: Maximum power savings
* Processor Power Management — Minimum Processor State — 5%
* Processor Power Management — Maximum Processor State — 50%
* Search and Indexing — Power savings mode: High Performance.
* Turn off the display: 20 minutes
* Adaptive display: On
* Multimedia settings — When sharing media: Prevent idling to sleep

Power Plan — High Performance

* Require password on: Yes
* Turn off hard disks after: 20 minutes
* Wireless adapter settings — Power Saving Mode: Maximum Performance
* Sleep — Sleep after: Never
* Sleep — Hibernate after: Never
* USB Settings — USB Selective suspend setting: Disabled
* Power button action: Shut Down
* Start menu power button action: Sleep
* PCI Express — Link State Power Management: Off
* Processor Power Management — Minimum Processor State — 100%
* Processor Power Management — Maximum Processor State — 100%
* Search and Indexing — Power savings mode: High Performance.
* Turn off the display: 20 minutes
* Adaptive display: Off
* Multimedia settings — When sharing media: Allow the computer to enter Away Mode

Source: http://www.energystar.gov/index.cfm?c=power_mgt.pr_power_mgt_winVista_default


And here is a good description for each settings from http://www.thehotfix.net/forums/ind...s-vista-power-management-guide-for-notebooks/

EDIT:

Thanks for the link. From what I've read, it seems that just the CPU is throttled, and the GPU is not affected.

 

I would like to see the tests conducted with a series of different benchmarks; I don't think we can go off of one benchmark.

I'll try PCMark Vantage and 3DMark Vantage on the machine I'm testing at the moment and report back later. The results could be interesting.

 

Since theres no difference in GPU's settings the difference will probably be on benchmarks that test the CPU (although both 06 and Vantage have a cpu score which is the one that we need to pay attention to verify this). Although i still dont get why it scores higher

I think theres something wrong with the program they benchmarked this to be honest.

 

That'd be great. I am waiting (b'marked this thread ). Thank you.

Unfortunately, I have no way to do these tests myself. I would love to do myself, though.. or at least be with and see in person when such tests are being done.. hmmmm.. hard life!!

 

Seems like balanced profile causes Cod4 to micro stutter in my case, which instantly stops when switching to High performance.

 

Update: CPU score in Vantage was 200~ higher in High performance profile. GPU score was the same.

 

I ran 3DMark06 on my test system and noticed a >2% difference between the High Performance and Balanced power profiles.

Alienware M17x; Specs: QX9300, dual GTX 280M SLI (179.85 drivers), 8GB DDR3-1333, 1TB RAID 0, Vista HP 64-bit

So, I'm going to have to nix that Balanced provides better performance. I can run other benchmarks if you wish. I don't have CrystalMark though.

 

Yup, I benchmakred my netbook and my desktop.

High performance is indeed a bit faster than balanced profile, the opposite of what the "controversy" described. The diference is about 1-2%.

 

Interesting. Well even if the results are almost the same, wouldn't this at least prove that it's better off to have computers running on Balanced?

Here's a good post by SHoTTa35 on neowin.net:

What do you think?


BTW, here's some more interesting results. Some people think that there's a bug in which switching power plans doesn't actually change the registry setting.
http://forums.worldofwarcraft.com/thread.html?topicId=5103774012&sid=1

http://support.microsoft.com/kb/935799

EDIT:

Thanks for the test. So is it most likely that CrystalMark 2004R3 used in her original tests was outdated and buggy?

 

I suggest leaving notebooks on Balanced . . . as noted, the CPU and GPU will clock as need to handle demand. The clock switching literally happens in thousandths of a second, so there won't be any performance loss.

 

Chaz on your 3dmark test the CPU score was almost the same. So CPU wise they had the same performance the problem is that your sm2.0 score dropped by 300 in one case which seems that the problem is somewhere else and not the power plan itself. Considering of course that power plan doesnt affect gpu performance (apart from the last option which should be the same in both cases).

 

Yes, but the difference was very small (~5%), and that can be attributed to benchmarking variability. I have run 3DM06 about 20 times on this laptop and that score is normal.

 

Yeap thats what i meant, it didnt really make any difference in 3dmark 06 (apart from the sm2.0 score which if you run the benchmark a few times you will probably get different scores) yey for benchmark programs stability .

Have to run some more tests on Vantage which i think its a bit more stable on its scores than 06, first time i run it balanced was 200 score bellow high performance which in my case was about 4%.

What worries me though, is it possible for whatever reason for the cpu to change multipliers during gameplay ? (not possible in theory since upon launching the game the multiplier should go on full speed) but with balanced profile i noticed some weird behaviour in some games, mostly micro stuttering in some cases.

 

very interesting findings. i had no idea. i feel stupid, which doesnt happen very often + rep. anywho im gonna test it myself on my laptop

 

The benchmarking scores do jump around quite a bit; I usually always run benchmarks for reviews 3+ times to ensure that I can duplicate them.

It is possible for the CPU to change multipliers during gameplay, but if the CPU load is relatively constant, then it should not. Even if it does, on paper it is unlikely to affect gameplay since the CPU can switch between multipliers/power states almost instantaneously.

Yes - if we could get more people to try this that would be helpful. Run one of the following:

• 3DMark06/Vantage
• PCMark05/Vantage

. . . and post your results using the Balanced and High Performance power profiles.

 

thanks. Looking forward to your test findings. I'll probably run some of my own tests soon as well.

We're having some mixed results but so far it seems that Balanced may be the preferred option because you can get the same performance while also saving some power.

I'm also wondering about how the 3rd party so-to-speak power options such as HP's recommended power setting compare to the Balanced in test results.
Maybe we could make our own NBR power option with the best possible results.

I noticed that it's more than just the CPU that could affect performance when switching from Balanced to High Performance or vica versa.

For example, PCI Express — Link State Power Management is set to Moderate power savings on Balanced and Off on High Performance.


And for anyone wondering what "Adaptive Display" is, here's a description I found from a forum:

----------------

Interesting stuff going on here. Looking forward to comparing more test results.

 

Scratch what i said about Vantage, just run the benchmark (cpu test only) about 10 times and the scores are pretty much identical just the score jumping 150~ up and down.

So basically the multiplier changes based on the cpu load ? Is there a program out there that can "record" the multiplier activity ? Im interested on checking if the cpu actually switches speed during games and if thats noticeable or not.

 

I want to try too, tonight Let me see how it goes.

 

Run some tests with wprime as well, results are pretty much the same.

 

Thanks for posting the results. So since they are the same, do you plan on using Balanced instead of High Performance?

BTW, I also think that Windows shouldn't make it seem that Balanced is not as powerful as High Performance.



For an average user, you see the dots and assume that Balanced isn't going to be as good as High Performance. In fact the dots make it seem that Balanced is only half as powerful. But from our tests, it seems that performance-wise they are almost identical. Windows can be very misleading here.

 

As far as i can tell theres no difference in performance from using balanced profile instead of high performance and you get the benefit that your computer is cooler when you are not doing anything intensive (and quitter most probably) since the cpu will downclock itself.

So yea balanced profile can/should "replace" the default windows profile IMO since almost everyone nowadays use high performance profile.

The only thing im trying to figure out is if Link state power management under PCI Express (which is set as moderate power savings by default in balanced profile) will cause any problems (performance wise).

True that but dave said something about win7 have high performance "hidden".

edit: Hm which im not sure if thats true since my win7 is showing this

 

I agree. These findings should show people that Balanced doesn't mean less performance.

And is there any way we can test the PCI Express so we can see how the two power options compare?

BTW, what about the power settings like HP recommended?

And regarding windows 7- I don't have a computer running it now, but according to this screenshot I found, it appears different than yours:



Maybe a different version of the release candidate? I wonder how the final release power settings will look.

 

Yea, my version is older than RC and RTM so they probably changed that.
Shouldnt matter what brand the notebook is.

Trying to throw a bit more benchmarks out there, a bit lazy at the moment so im working with what i have instead of installing games to benchmark them

CSS

 

Interesting. In that test, Balanced had a higher fps than High Performance. Although it's minor, it once again proves that High Performance isn't necessarily better.

Besides benchmark tests, how does the Balanced power plan compare to High Performance as far as your own opinion doing your everyday tasks?
Also, when in Balanced, is the CPU throttled down during regular web browsing?

 

In my case the CPU downclocks to 1.6ghz (from 2.8) but clicking a url or opening a program/folder/anything in general instantly boosts the cpu back to 2.8ghz till it loads/launches etc. So i doubt it will affect everyday tasks in any way.

Even clicking urls takes the cpu back to 2.8 (or 2133mhz) for 1-2 secs (page loading).

 

I see. So while you're just viewing a page, the CPU downclocks, therefore, saving power. And once you click a link, the CPU goes back to full potential in order to load the page as fast as possible.
So it sounds like it'll help keep the computer cooler too.

 

Yeap exactly

 

I got my first laptop a few weeks ago and have been running it on High Performance when it is plugged in... Seeing this thread though I have just converted to using the Balanced option, especially since I really do not do any intensive tasks on my laptop...

And those dots are misleading because I definitely thought that High Performance was much better

 

Yep, I have to agree with the above results and posts - got similar marks with 3Dmark and wPrime with balanced and high performance.. well well.. this has got to be my favorite thread here, these days

 

Thanks for your results. So it seems pretty unanimous that "Balanced" is the way to go.
I'm glad I did some research on this because it seems that most people took for granted that "High Performance" is the best, but these results prove that wrong.
This is somewhat of a ground-breaking discovery in the world of Windows.


BTW, I too have made the switch to Balanced and if anything, it seems faster. Plus I'm glad to know that I'm stressing my computer less when idling.

 

Excellent thread! I always had performance set to HIGH thinking it offered faster performance. although I don't play any games.....heh

quick question.....

what do u guys think of the wireless adapter power saving?

I always have it on high performance as I am usually in cafes and I would like to catch the farthest available network possible if needed.

would I still be able to catch the same network if i was on power saving mode for the wireless adapter settings? or does that performance not have anything to do with the distance of networks that i can connect to?

 

I think if you want to be able to achieve the maximum distance possible out of your wi-fi, you should have the Wireless Adapter settings on Max Performance. But if you really want to get extra power savings, then try connecting to the wireless network and then switching to Power Saver with the wireless adapter on max power saving and see if you maintain connection.

EDIT:

Here's a good description I found:

 

High Performance still throttles the CPU and GPU (speedstep) but i have a feeling it disables the CPU sleep states (C4, C4E, etc.).

 

Interesting...Is there any way to test and see the CPU sleep states so we can tell if they change when we go from "High Performance" to "Balanced"?

What about this:

In High Performance, his CPU didn't downclock, but when he used "Balanced", then the changes he mentioned happen. BTW, what program did you use CooLMinE to see the clock speed of the CPU?


And here's the info regarding the sleep states according to intel:

 

And regarding the Battery Life between the power plans- here's a test that was run by Smart Computing:

Have you tried using CPU-Z ?


Also, I found a very interesting thread regarding how the Power Plans affect our CPUs in the real world.

Here's the link: http://blogs.msdn.com/richardt/archive/2007/07/11/vista-power-management-cpu-frequency.aspx

But basically these charts explain it:

NOTE: the blue line shows you the CPU frequency - the green line shows you CPU utilization

High Performance:



Balanced:



Power Saver:

 

Im using rainmeter and samurize, both sit on my desktop and display the changes to the cpu clock/freq (among other things as well ).

Using high performance profile you see the CPU clock staying at 2.8 and not bunching from there, ever.

If you switch to balanced it instantly clocks down to 1.6 and switches automatically depending on what you are doing.

Basically the info from the resource overview above show exactly the same thing.

Sorry for the late reply, sleep + work = a nice 16 wasted hours

 

Thanks for the info. So the charts I posted above seem to be just like you mentioned. So Balanced is the best way to go since it will work as hard as High Performance when needed, but save power when there's little activity.

We need to get the word out that Balanced is the best choice.

 

Thanks to this thread, I now have 50% more battery life, with the same performance!!! I LOVE the OP!

 

So far I've been running my computer on balanced for a few days now and haven't noticed any slowdown. My computer does seem to be running cooler too and the fans aren't as active.

I'm still opening to anyone proving otherwise, but as of now, it seems that there's no reason to use the "High Performance" power plan.

I dont' want benchmarks to be our only means of judging this. So for those of you that have switched to Balanced, how do feel the performance of your computer has been since?

 

I don't get it. Why is it surprising that it "works" the way it should be? It seems people here have been fooled a lot lately to even believe their own eyes.

 

Mostly cause of the name (Balanced,High performance) and the dots next to the profiles M$ used

Most if not all gamers nowadays seem to use high performance cause they think that it works better than balanced (for the reasons i said above and probably a few more).

So actually staring the facts makes you realize how wrong you were

 

Ahh, I see.

I can only assume "micro-stutters" happen because while fast in our standards, the switching speed is still not fast enough. Ideally in games the CPU should clock down only when the CPU part of the job is done before its sent to the GPU for actual rendering.

This should be fixed with i7.

Update: I think I get it: "High Performance still throttles the CPU and GPU (speedstep) but i have a feeling it disables the CPU sleep states (C4, C4E, etc.)."

See, the CPU has multiple speed grades and sleep states it can switch to. The state it ends up is determined by how long will the program will allow the CPU to stay at idle. Each lower power steps take subsequently longer time to go back to the full speed. While clock speed changes might be fast enough not to be noticeable, coming out of each sleep state could be taking enough time to notice that brief stutter.